正点原子RK3588（一）——开机测试+AI初探

一、adb

adb shell，进入板子的根目录
exit，退出到linux
adb pull 板子 linux（从板子到linux）
adb push linux 板子（从linux到板子）

二、测试AI功能

2.1 resnet18

import cv2
import numpy as np
import platform
from synset_label import labels
from rknnlite.api import RKNNLite# decice tree for RK356x/RK3576/RK3588
DEVICE_COMPATIBLE_NODE = '/proc/device-tree/compatible'def get_host():# get platform and device typesystem = platform.system()machine = platform.machine()os_machine = system + '-' + machineif os_machine == 'Linux-aarch64':try:with open(DEVICE_COMPATIBLE_NODE) as f:device_compatible_str = f.read()if 'rk3562' in device_compatible_str:host = 'RK3562'elif 'rk3576' in device_compatible_str:host = 'RK3576'elif 'rk3588' in device_compatible_str:host = 'RK3588'else:host = 'RK3566_RK3568'except IOError:print('Read device node {} failed.'.format(DEVICE_COMPATIBLE_NODE))exit(-1)else:host = os_machinereturn hostINPUT_SIZE = 224RK3566_RK3568_RKNN_MODEL = 'resnet18_for_rk3566_rk3568.rknn'
RK3588_RKNN_MODEL = 'resnet18_for_rk3588.rknn'
RK3562_RKNN_MODEL = 'resnet18_for_rk3562.rknn'
RK3576_RKNN_MODEL = 'resnet18_for_rk3576.rknn'def show_top5(result):output = result[0].reshape(-1)# Softmaxoutput = np.exp(output) / np.sum(np.exp(output))# Get the indices of the top 5 largest valuesoutput_sorted_indices = np.argsort(output)[::-1][:5]top5_str = 'resnet18\n-----TOP 5-----\n'for i, index in enumerate(output_sorted_indices):value = output[index]if value > 0:topi = '[{:>3d}] score:{:.6f} class:"{}"\n'.format(index, value, labels[index])else:topi = '-1: 0.0\n'top5_str += topiprint(top5_str)if __name__ == '__main__':# Get device informationhost_name = get_host()if host_name == 'RK3566_RK3568':rknn_model = RK3566_RK3568_RKNN_MODELelif host_name == 'RK3562':rknn_model = RK3562_RKNN_MODELelif host_name == 'RK3576':rknn_model = RK3576_RKNN_MODELelif host_name == 'RK3588':rknn_model = RK3588_RKNN_MODELelse:print("This demo cannot run on the current platform: {}".format(host_name))exit(-1)rknn_lite = RKNNLite()# Load RKNN modelprint('--> Load RKNN model')ret = rknn_lite.load_rknn(rknn_model)if ret != 0:print('Load RKNN model failed')exit(ret)print('done')ori_img = cv2.imread('./space_shuttle_224.jpg')img = cv2.cvtColor(ori_img, cv2.COLOR_BGR2RGB)img = np.expand_dims(img, 0)# Init runtime environmentprint('--> Init runtime environment')# Run on RK356x / RK3576 / RK3588 with Debian OS, do not need specify target.if host_name in ['RK3576', 'RK3588']:# For RK3576 / RK3588, specify which NPU core the model runs on through the core_mask parameter.ret = rknn_lite.init_runtime(core_mask=RKNNLite.NPU_CORE_0)else:ret = rknn_lite.init_runtime()if ret != 0:print('Init runtime environment failed')exit(ret)print('done')# Inferenceprint('--> Running model')outputs = rknn_lite.inference(inputs=[img])# Show the classification resultsshow_top5(outputs)print('done')rknn_lite.release()

在这里插入图片描述

2.2 对于安装

1.安装anacoda
2. 安装adb
3. 安装交叉编译工具链
4. 安装RKNN Toolkit Lite2
5. 更新NPU驱动（待）
6. 查看和调整CPU与NPU频率（待）
7. 安装rknn-toolkit2（待）
8. 安装conda（待）

RKNN Toolkit Lite2和rknn-toolkit2
rknn-toolkit2 的出现解决了这一难题。它提供了一个完整的解决方案，允许开发者将已经训练好的深度学习模型（如传统的TensorFlow、Pytorch、onnx 等格式的模型）转换为优化的rknn 格式，这些格式的模型能够充分利用RK3568、RK3588 等设备的硬件特性，从而在保证模型性能的同时，降低了对计算资源的需求。（优化，变成RKNN）

RKNN Toolkit Lite2 是一个轻量级的工具，专门为带有Rockchip NPU平台提供Python编程接口，帮助用户部署使用RKNN-Toolkit2导出的RKNN模型。它更侧重于在开发板上的部署和推理，通常用于实际的硬件环境中。RKNN Toolkit Lite2的开发文档位于 external/rknn/rknn-toolkit2/rknn-toolkit-lite2/doc 目录下，并且在Toybrick Debian11系统中已经预装了RKNN Toolkit Lite2的whl包。

rknn-toolkit2 是一个更全面的开发套件，它不仅包括了Lite2的功能，还提供了更多的工具和接口，用于在PC平台上进行模型转换、推理和性能评估。它支持的功能更多，包括模型转换、量化、性能和内存评估等。rknn-toolkit2的开发文档位于 external/rknn/rknn-toolkit2/doc 目录下。

本文来自互联网用户投稿，该文观点仅代表作者本人，不代表本站立场。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如若转载，请注明出处：http://www.xdnf.cn/news/145843.html

如若内容造成侵权/违法违规/事实不符，请联系一条长河网进行投诉反馈，一经查实，立即删除！